Much of the original U.S. grassland has undergone conversion to cropland. During the last few years, large first- and second-order watershed scale projects have begun to reconstruct the native tallgrass prairie cover and biodiversity. The effect on watershed hydrological budget is largely unknown, especially concerning storm run-off. Curve number variability is used to estimate the uncertainty of peak run-off following the change in cover, given rainfall recurrence and watershed size. The method involves three steps: (1) estimate the time-of-concentration for many similar sized watersheds in the region, (2) define the probability distribution for time-of-concentration, curve numbers, and watershed area, (3) with these data, generate input variables for a Monte Carlo analysis, which can then be used to predict the mean and confidence interval of peak run-off. As an example, spatial and hydrological characteristics of first- and second-order watersheds ranging from 2 to 50 km2 in the Red River of the North basin provide a log normal probability distribution for time-of-concentration. Using the range of watershed area and a β probability distribution for curve number uncertainty, the analysis predicts the change in peak run-off from an ensemble of watershed realizations that characterize cropland to grassland conversion. Results suggest that given five- and 25-year, 24-hour rainfall recurrence, average reduction in peak run-off will range from 50 to 55% and 40 to 45%, respectively, for the basin. A large range of uncertainty at the 80% confidence interval, however, indicates that an accurate prediction requires analysis for specific watersheds.